Hu Zhengming is not only good at scientific research in the field of integrated circuits, but is also a top scientist in this field. He also has a deep understanding of the chip industry and his own thinking.
Because Hu Zhengming studied for a doctorate at Berkeley in 1969, went to MIT to work as an assistant professor for just three or four years, and then returned to Berkeley. It can be said that he spent half of his life in Berkeley and witnessed the development of Silicon Valley with his own eyes.
The early history of Silicon Valley cannot be bypassed. Fairchild Semiconductor was once the largest semiconductor manufacturer in the world. In the year before 1969, all eight scientists on the founding team of Fairchild Semiconductor left Fairchild, thus starting Silicon Valley. entrepreneurial boom.
In the year when Hu Zhengming arrived at Berkeley, at the Semiconductor Engineers Conference held in Sunnyvale, there were less than 24 engineers among the 400 attendees who had not worked at Fairchild.
Hu Zhengming watched Silicon Valley's transformation from the semiconductor boom to the rise of personal PCs to the Internet wave in the past decade, but his focus was more on the changes brought about by the advancement of underlying technologies.
Zhou Xin said: "I know that Dr. Yu Youcheng joined CAAS after the establishment of CAAS, and I have always been in contact with him.
I have some understanding of TSMC's development path. China's semiconductor industry wants to catch up, and cooperation with Intel is the fastest and most convenient way. "
Yu Youcheng is the vice president of Intel and has a PhD in electrical engineering from Stanford. He has joined Intel since 1972. Yu Youcheng was recruited by the aforementioned Intel Chairman Andy Grove.
Yu Youcheng left Intel in 1977 to form a company called Video Brain, focusing on the manufacturing of home computers. At that time, he founded Video Brain because he believed that computers would enter homes sooner or later. However, due to the wrong timing of its launch, it eventually went bankrupt, and Yu Youcheng returned to Intel.
Yu Youcheng, like Hu Zhengming, was born and raised in mainland China, then went to America to study and stayed in America. Many engineers in the semiconductor field followed this growth path. If the mainland had the same conditions, it would have been able to produce a large number of talents in the integrated circuit field in that era.
Hu Zhengming: "Unlike the Internet industry, the development of the integrated circuit field cannot rely on the ingenuity of a few people. It requires many talented engineers to burst out sparks of inspiration in each subdivision, and then ultimately promote technology through hard work. improvement.
Even though China, as a pursuer, has a clear path ahead, and all technical principles can be found in public information, it is still not that simple to achieve success in the field of integrated circuits.
In fact, I have never been worried about your talent. I am just worried that you will succeed too easily and too quickly, and underestimate the difficulty of development in the field of integrated circuits.
Do you know why I never promised you? I know I helped you with your work, but it didn't work out in the end. You certainly couldn't bear to see me, a teacher, starve to death. "
Hu Zhengming showed a sly smile when he said the last sentence. Zhou Xin thought about it and then said, "Is it because the difficulty is too high?"
Hu Zhengming nodded: "Yes, the difficulty is far beyond your imagination.
The integrated circuit, that is, the chip industry, is different from all previous industries, because the chip industry is a civilian industry, and the most important thing is profitability.
Technical feasibility is secondary, and commercial profitability is the essence of this matter.
And this kind of profit not only means that the company wants to make a profit, but also means that the company sells its products to other companies, and other companies can also make profits by using your products.
Why would Nikon be willing to package and sell you the lithography machine technology that is two generations behind? Because Nikon knows that even if you build the lithography machine, no manufacturer will choose to use the new core lithography machine.
The earliest commercial lithography machine was Amerikan's GCA company. At that time, it was probably an exposure machine to be precise. In the 1960s, GCA was able to sell dozens of exposure machines every year.
Later, GCA began to ask Nikon for help, and Nikon became GCA's lens supplier. Philips Physical Laboratory, the predecessor of ASML, turned to Zeiss as their lens supplier. In the early 1980s, Philips' photolithography machines were still in their infancy, but GCA's annual sales had exceeded US$100 million, a tenfold increase in three years.
Not a single lithography machine in Philips Physics Laboratory was sold, and then the board of directors of Philips Semiconductor planned to sell Philips Physics Laboratory. ASM originally wanted to buy it, but Philips felt that ASM was too small and could not afford the price.
Finally, under the leadership of Cruyff, the technical director in charge of the lithography machine business, Philips and ASM jointly established the current ASML. Cruyff and I usually chat at the annual IEEE. He has a superb understanding of technology. With his unusually keen sense of smell, he keenly discovered the great prospects of photolithography machines.
Nikon was originally a supplier of GCA, but later discovered that there was a leader in the field of photolithography machines, and they also began to enter this field. Backed by the neon chip industry chain, Nikon took off rapidly. It only took five years to go from scratch to competing with GCA in the lithography machine market.
Nikon and Canon, which rely on Neon's chip industry chain, can develop by relying on Neon's chip industry chain.
However, ASML has technical advantages. The photolithography machines of Nikon and GCA have an accuracy of more than 1 micron. They use lead screws to move the wafer stage. The accuracy of image detail positioning will be greater than 1 micron. ASML can achieve less than 1 micron. With the help of this technical advantage, ASML was able to overtake and catch up. But even in the early days, ASML still relied on R&D grants and subsidies from the Dutch government to survive.
Later, it was not until TSMC, IBM, and Samsung made ASML their suppliers that ASML's technical advantages began to come into play.
As for how GCA was surpassed by Nikon and Canon, the strength of the Neon semiconductor industry chain is one reason. After Nikon entered the field of photolithography machines, GCA changed their lens supplier to Zeiss. The problem with Zeiss’ lenses is another reason. Because of the combination of the two, GCA was quickly squeezed out of the market by Nikon and eventually went bankrupt.
What Nikon is facing now is the same dilemma that GCA faced at the time. Their market in photolithography machines is constantly being squeezed by ASML, and profits and revenue are declining. Nikon will sell you their backward technology because ASML puts too much pressure on them.
Selling backward technology to Xinxin can make the financial reports of the lithography machine business look better. The more important reason is that Nikon does not believe that Chinese companies can pose any threat to them in the field of lithography machines.
From the market level, China's chip industry chain is extremely weak, and it cannot support China's photolithography machines to enter the production process. Failure to enter the production process means that there will never be room for progress, and technological progress cannot be produced.
From a technical perspective, China does not have enough talent reserves. Nikon does not believe that China has unique technology similar to ASML's breakthrough.
Lithography machines are only one link in the chip industry chain. What about other links? Hua Guo has too many lessons to make up for.
I promised you, but I need time to think carefully about where to start after I go to form a positive cycle as soon as possible.
Profitability is too important for the chip industry. SMIC, which is also a Chinese company, will not use Xinxin lithography machines. If I were the person in charge of SMIC, I would not consider it either.
Why? Using Nikon or ASML equipment, the yield rate is above 95%. If there is any problem with the equipment, the engineers from the original lithography machine factory can help me repair it in three days or even less. If I use a new core lithography machine, The yield rate may only be 80%, and it will take slow running-in and adjustments to increase the yield rate to 95%. Once the equipment fails, the time required to repair it is much longer than that of products from mature manufacturers.
This in turn is related to the profitability of chip foundry manufacturers. Using Nikon's photolithography machines, the profit margin is 50%. Taking into account product depreciation and capital costs, the profit of chip foundry is very thin, less than 10%. Why is TSMC's market value so high? It's because they have developed economies of scale. Even including equipment depreciation and capital costs, the profit margin can reach 14%. Last year, it seemed that it was only 10%.
When using new core lithography machines, excluding product depreciation and capital costs, manufacturers will even lose money, which is obviously unacceptable to them. "
This is also the fundamental reason why China’s semiconductor industry will never develop if Amerikan does not take a series of measures to stop it.
Foreign equipment is so easy to use, who would use domestic equipment? Not to mention that the semiconductor field is all private enterprises and not state-owned enterprises. There is no such thing as a political mission. Everyone is pursuing profits.
Americen's approach has given the domestic semiconductor industry chain a glimmer of hope.
Zhou Xin said: "I know that Huahong Semiconductor is currently doing prototype testing of Xinxin lithography machines. They are a state-owned chip manufacturing company in China. Testing Xinxin lithography machines is a political task for them, so they can promote it.
But there is also an exchange of interests, because Huahong Semiconductor's chip production line was built with Neon's help, and they originally used Nikon's lithography machines, so it was very easy to get started when testing new core lithography machines.
Therefore, the cooperation between Xinxin Lithography Machine and Huahong Semiconductor has natural advantages. Moreover, the process of the new core lithography machine is 220nm, while Huahong Semiconductor's most advanced process is only 0.35 micron. This means that Huahong can help us test the lithography machine while also promoting its technological progress.
Moreover, Xinxin Technology also has natural advantages in cooperating with Huahong as a whole. .
Professor, do you know China’s 908 Project and 909 Project? "
Hu Zhengming fell into memories. He was ten years older than Yang Chonghe. Yang Chonghe had a long chat with him before returning to China. The conversation was about why Project 908 failed and how to promote China's chips after returning to China. develop.
(Yang Chonghe mentioned in Chapter 60, the founder of Xintao Technology and Montage Technology)
Hu Zhengming has stayed in Silicon Valley for so many years, and Yang Chonghe is one of the first semiconductor people to return to China. In this era, it is difficult to get a doctorate degree, which is very valuable. Returning to China is purely to open up wasteland. Not many people are willing to go back. Most of them return to China in the end.
Therefore, Hu Zhengming was deeply impressed by Yang Chonghe.
"I know that the final result of Project 908 was not very satisfactory. It can be said to be a failure. It seems that they lost 240 million yuan in the first year. They were too slow. It took seven years from project establishment to production. In these seven years Times have changed in the chip industry, and the situation is completely different from the situation during the pre-research period.
Later, Project 909 was launched. Project 909 did not adopt the same vertically integrated manufacturing model as 908, but adopted a spin-off model, that is, separating chip foundry and chip design. The reason why this model was chosen was precisely because Yang Chonghe had a long conversation with the then Minister of Electronic Industry and persuaded him.
Yang Chonghe seems to have established a chip design company and received investment from a group of investment companies in Silicon Valley, Neon and Wanwan. "
The company founded by Yang Chonghe is called Xintao. It is the first chip design company in China to receive orders from developed countries. In 1999, they received an order from Panasonic.
Zhou Xin said: "The Xintao company he founded has been acquired because there were too many investors behind it and the founder team lost control. It was eventually acquired by Amerikan IDT."
The acquisition price of IDT was US$85 million, and venture capital institutions made a lot of money.
Zhou Xin continued: "When Xintao was acquired, Uncle Guan wanted to poach Yang Chonghe. Uncle Guan believed that Yang Chonghe was very capable and was a rare talent in China who understood both technology and management.
But unfortunately, the agreement signed by Yang Chonghe restricted him to work for IDT for three years, which means that he could join Xinxin Technology after three years. "
After hearing this, Hu Zhengming sighed: “So venture capital has good and bad qualities, it’s a double-edged sword.
Sometimes taking too much money from venture capital is not necessarily a good thing.
Now that I think about it, it was because I took too much money from your investment company that I spent the last part of my career working for you in Shenhai. "Hu Zhengming finally joked.
Zhou Xin said: "Professor, if you don't want to, I can't force you.
Let’s go back to Project 909. Project 909 is like what you just said. Huahong Semiconductor is a chip foundry, that is, a manufacturer. They used to mainly make memory chips.
You know that from the second half of last year to the first half of this year, as the Nasdaq bubble burst, it led to a depression in the semiconductor industry.
This has led to low-priced dumping of chips from Amerikan, Neon and Koryo, and the price of memory chips is less than one-tenth of the previous price. Hua Hong Semiconductor lost 700 million yuan in the first half of this year. "
A loss of 700 million in half a year is obviously unacceptable to Hua Hong Semiconductor. It is even more unacceptable for the 909 project with a national investment of 10 billion yuan to fail again.
Later, Huahong withdrew from the memory chip field, and then Yangtze Memory returned to the memory chip field. This is a reincarnation. If you don’t persevere then, you will have to pay a higher price later.
To be honest, Huahong Semiconductor's huge losses are also caused by Neon Digging. Neon took foreign exchange from China to help build chip production lines, and then guaranteed to underwrite the products.
However, Neon's conditions for underwriting products are to underwrite products at market prices, and because of its cooperation with Neon, memory chips have become Huahong Semiconductor's only product. The price of memory chips has dropped from US$20 to US$2. Huahong regrets that it is too late. .
(End of chapter)
